From 909a32c82eec2bfd3243d122ae81553242f3502a Mon Sep 17 00:00:00 2001
From: serge-sans-paille <sergesanspaille@free.fr>
Date: Sat, 1 Nov 2025 00:13:23 +0100
Subject: [PATCH 1/2] Fix widen_t signed case

---
 include/xsimd/types/xsimd_traits.hpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/xsimd/types/xsimd_traits.hpp b/include/xsimd/types/xsimd_traits.hpp
index 1e1e6bc77..20c97f89f 100644
--- a/include/xsimd/types/xsimd_traits.hpp
+++ b/include/xsimd/types/xsimd_traits.hpp
@@ -336,8 +336,9 @@ namespace xsimd
     namespace detail
     {
         template <typename T>
-        struct widen : widen<typename std::make_unsigned<T>::type>
+        struct widen
         {
+            using type = typename std::make_signed<typename widen<typename std::make_unsigned<T>::type>::type>::type;
         };
 
         template <>

From 1d4bd9cd3fbb10c0c5f2b44700e82a27107a3e3c Mon Sep 17 00:00:00 2001
From: serge-sans-paille <sergesanspaille@free.fr>
Date: Fri, 31 Oct 2025 23:24:33 +0100
Subject: [PATCH 2/2] Add support for xsimd::widen on ARM

Related to #1179
---
 include/xsimd/arch/xsimd_neon.hpp   | 36 +++++++++++++++++++++++++++++
 include/xsimd/arch/xsimd_neon64.hpp |  9 ++++++++
 2 files changed, 45 insertions(+)

diff --git a/include/xsimd/arch/xsimd_neon.hpp b/include/xsimd/arch/xsimd_neon.hpp
index ad2a25246..9ddde6b78 100644
--- a/include/xsimd/arch/xsimd_neon.hpp
+++ b/include/xsimd/arch/xsimd_neon.hpp
@@ -20,6 +20,7 @@
 
 #include "../types/xsimd_neon_register.hpp"
 #include "../types/xsimd_utils.hpp"
+#include "./common/xsimd_common_cast.hpp"
 
 // Wrap intrinsics so we can pass them as function pointers
 // - OP: intrinsics name prefix, e.g., vorrq
@@ -3211,6 +3212,41 @@ namespace xsimd
         {
             return vreinterpretq_f32_u32(swizzle(batch<uint32_t, A>(vreinterpretq_u32_f32(self)), mask, A {}));
         }
+
+        /*********
+         * widen *
+         *********/
+        template <class A, class T, detail::enable_sized_signed_t<T, 1> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_s8(vget_low_s8(x), vdup_n_s8(0))), batch<widen_t<T>, A>(vaddl_s8(vget_high_s8(x), vdup_n_s8(0))) };
+        }
+        template <class A, class T, detail::enable_sized_unsigned_t<T, 1> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_u8(vget_low_u8(x), vdup_n_u8(0))), batch<widen_t<T>, A>(vaddl_u8(vget_high_u8(x), vdup_n_u8(0))) };
+        }
+        template <class A, class T, detail::enable_sized_signed_t<T, 2> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_s16(vget_low_s16(x), vdup_n_s16(0))), batch<widen_t<T>, A>(vaddl_s16(vget_high_s16(x), vdup_n_s16(0))) };
+        }
+        template <class A, class T, detail::enable_sized_unsigned_t<T, 2> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_u16(vget_low_u16(x), vdup_n_u16(0))), batch<widen_t<T>, A>(vaddl_u16(vget_high_u16(x), vdup_n_u16(0))) };
+        }
+        template <class A, class T, detail::enable_sized_signed_t<T, 4> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_s32(vget_low_s32(x), vdup_n_s32(0))), batch<widen_t<T>, A>(vaddl_s32(vget_high_s32(x), vdup_n_s32(0))) };
+        }
+        template <class A, class T, detail::enable_sized_unsigned_t<T, 4> = 0>
+        XSIMD_INLINE std::array<batch<widen_t<T>, A>, 2> widen(batch<T, A> const& x, requires_arch<neon>) noexcept
+        {
+            return { batch<widen_t<T>, A>(vaddl_u32(vget_low_u32(x), vdup_n_u32(0))), batch<widen_t<T>, A>(vaddl_u32(vget_high_u32(x), vdup_n_u32(0))) };
+        }
+
     }
 
 }
diff --git a/include/xsimd/arch/xsimd_neon64.hpp b/include/xsimd/arch/xsimd_neon64.hpp
index ca7512e38..7cb531be9 100644
--- a/include/xsimd/arch/xsimd_neon64.hpp
+++ b/include/xsimd/arch/xsimd_neon64.hpp
@@ -1547,6 +1547,15 @@ namespace xsimd
         {
             return batch<std::complex<double>>(swizzle(self.real(), idx, A()), swizzle(self.imag(), idx, A()));
         }
+
+        /*********
+         * widen *
+         *********/
+        template <class A, class T>
+        XSIMD_INLINE std::array<batch<double, A>, 2> widen(batch<float, A> const& x, requires_arch<neon64>) noexcept
+        {
+            return { batch<double, A>(vcvt_f64_f32(vget_low_f32(x))), batch<double, A>(vcvt_high_f64_f32(x)) };
+        }
     }
 }